Computational modelling of LY303511 and TRAIL-induced apoptosis suggests dynamic regulation of cFLIP

Motivation: TRAIL has been widely studied for the ability to kill cancer cells selectively, but its clinical usefulness has been hindered by the development of resistance. Multiple compounds have been identified that sensitize cancer cells to TRAIL-induced apoptosis. The drug LY303511 (LY30), combined with TRAIL, caused synergistic (greater than additive) killing of multiple cancer cell lines. We used mathematical modelling and ordinary differential equations to represent how LY30 and TRAIL individually affect HeLa cells, and to predict how the combined treatment achieves synergy. Results: Model-based predictions were compared with in vitro experiments. The combination treatment model was successful at mimicking the synergistic levels of cell death caused by LY30 and TRAIL combined. However, there were significant failures of the model to mimic upstream activation at early time points, particularly the slope of caspase-8 activation. This flaw in the model led us to perform additional measurements of early caspase-8 activation. Surprisingly, caspase-8 exhibited a transient decrease in activity after LY30 treatment, prior to strong activation. cFLIP, an inhibitor of caspase-8 activation, was up-regulated briefly after 30 min of LY30 treatment, followed by a significant down-regulation over prolonged exposure. A further model suggested that LY30-induced fluctuation of cFLIP might result from tilting the ratio of two key species of reactive oxygen species (ROS), superoxide and hydrogen peroxide. Computational modelling extracted novel biological implications from measured dynamics, identified time intervals with unexplained effects, and clarified the non-monotonic effects of the drug LY30 on cFLIP during cancer cell apoptosis. Supplementary information: Supplementary data are available at Bioinformatics online. Contact: LisaTK@nus.edu.sg or Shazib_Pervaiz@nuhs.edu.sg

. Parameters 71-10 were estimated using our previous experimental work  as detailed below.

Supplementary Text 1.4: Parameter Estimation for the LY30 + TRAIL Model
LY30-induced Receptor Oligomerization. Quantification of receptor oligomerization was conducted using our previous experiments that pulled down DR5 from HeLa lysate using excess or limited amounts of antibody . The total amount of DR5 protein was the same in LY30-treated cells and untreated, but the amount of DR5 pulled down by limited antibodies was 5.8-fold greater in LY30-treated cells than in untreated cells ( Supplementary Figure 1.4.1). If the pull-down ratio had been close to 3-fold, we might have characterized the transition as trimerization, because trimerization is receptor stoichiometry for activation. Instead we considered analogies to published results finding that clustered or oligomerized or lipid-raft-localized receptors transduce TRAIL signals more readily than non-clustered receptors (Mellier and Pervaiz, 2012). For example, the DR clustering induced by LY30 resembles the lipid raft localization of DRs, induced by the ROS-producing compound resveratrol (Delmas, et al., 2004). Therefore, we modeled DR5 oligomerization as a change of state. In untreated cells, all receptors were assumed to be initially an unprimed state R. We then modeled that receptor interaction with LY30 would trigger a state transition towards a cluster-localized state called R_primed. The rate parameters of this reaction were chosen so that an initial dose of R would reach steady state levels of R_primed at approximately 1 hour, and also so that the R:LY30 complex would have low concentration. Experimental results reprinted from  and (b) quantified density of the DR5 bands using ImageJ. Was set to 1e-3, the same as Albeck et al. We used the following to estimate the final degree of freedom, the KK40 reaction for activation of the primed receptors After 5 minutes of exposure to TRAIL, immuno-precipitation of the FADD-containing complex with DR5 (the DISC-like complex) showed roughly 7-fold more assembly of DISC-like complexes in cells pre-treated with LY30, compared with cells that had not been pretreated. See Supplementary Figure 1 HeLa cells. HeLa cells were exposed to TRAIL (20ng/ml) for 5 min with or without prior addition of LY30(25µM) for one hr (Poh, et al.). (b) Quantification of FADD from Western blots.
The existing model was simulated to obtain the level of R* at 5 minutes after treatment with TRAIL alone. The experimentally measured fold-change in DISC-like complex formation at 5 minutes was used to compute an expected fold-change increase in R* for simulations of the LY+T treatment at 5 minutes. Then the KK40 parameter was adjusted so that simulations of LY+T would achieve the expected value of R* at 5 minutes.
Doses. We stimulated the model with 1200 molecules/cell of TRAIL to represent 20ng/ml of TRAIL, consistent with prior use of 3000 molecules/cell to represent 50ng/ml of TRAIL. The dose of LY30 was not calibrated to physical units, and the dose-response of LY30 was not studied. Therefore, simulations of LY30 should either use zero dose, or else use the same unitless dose (4.1) that we used when calibrating the simulated downstream effects of LY30 on receptors and cFLIP.

LY30-induced downregulation of cFLIP.
The phenomenon of cFLIP down-regulation upon LY30 treatment was modeled initially (prior to Fig.7)   .
Caspase Rates. Non-specific reactions for synthesis and degradation were added manually with the rates specified. Then automated parameter optimization was performed with the KroneckerBio Toolbox, varying the reaction rates for the caspase enzymes, so that simulations of LY30 alone and simulations of TRAIL alone would match previously published time kinetics of Caspase-3 and Caspase-8 activity after treatment with LY30 alone and TRAIL alone .

LY30-induced ROS production.
As we did not know the exact mechanism behind the ROS production upon LY30 treatment, so we simulated a simplistic pseudo-reaction in which LY30 induces ROS production, ROS is an ineffective intermediate state and ROS* represents effective ROS that can contribute to apoptosis. The reaction rates for ROS to cause apoptosis or mitochondrial permeabilization were adjusted manually, to maximize the agreement between the Monte Carlo population simulation of viability, and the observed viability, as a fold-change from untreated to LY30-treated conditions.

Supplementary Figure 1.5: Single Cells versus Cell Populations (in silico)
Supplementary Figure 1.5: Comparison of simulation profiles between single cells and a cell population. The x-axis is time after adding drug, and the y-axis is the relative concentration of cPARP. In this figure, each solid line represents one single cell and the dotted line with orange filing is the averaged result of 10,000 single cells.

Supplementary Text 1.6: Conversion of absolute activity into relative foldchange
The model is simulated using absolute protein levels, but Figure 4 shows simulations plotted as predictions of experimental observations in terms of relative enzyme activity. The caspase family of cysteine proteases can target different substrates according to structure. Caspase-8 activity is typically measured as the cleavage of the peptide IETD (Ile-Glu-Thr-Asp), using the fluorescence of Ac-IETD-AFC. Caspase-3 activity is typically measured as the cleavage of DEVD, using the fluorescence of Ac-Asp-Glu-Val-Asp-pNA. Fluorescent substrates were purchased from (BioMol, Plymouth Meeting, PA, USA) and used according to manufacturer's instructions. McStay et al. found that commercial caspase substrates, such as the ones we used, are not entirely specific for their respective cysteine proteases (McStay, et al., 2008), and there is overlapping cleavage motif selectivity. We converted our simulations of absolute protein levels into hypothetical IETD and DEVD fluorescence using a weighted combination of individual caspase levels, according to the quantification of overlap in (McStay, et al., 2008). IETD_abs (DEVD_abs) refers to absolute IETD (DEVD) fluorescence and IETD_rel (DEVD_rel) refers to relative IETD (DEVD) fluorescence in treated cells versus untreated cells.

Supplementary Figure 1.7: Caspase-3 activity in experiments and simulations.
Model simulation was converted into relative activity using the equations in Supplementary Text 1.1.5. Note that simulations of Caspase-3 would not be expected to agree with experiments if there are flaws in the model simulation upstream of Caspase-3, as described in section 3.5 of the main text.

The ROS-cFLIP Model
Supplementary This ROS-cFLIP model is intended to represent multiple possible sources of ROS including NADPH oxidase. However, for specificity of kinetics, the reactions describe ROS production by mitochondrial respiratory chain complexes (Buettner, et al., 2006).

Supplementary Figure 4.7: FADD expression.
Supplementary Staining with DCFDA to indicate ROS in HeLa cells treated with 25uM LY30 for 5 or 10 minutes in the presence or absence of ROS scavengers. 10mM Tiron was pre-incubated for 2 hours before the addition of LY30 while 2000U/mL catalase was pre-incubated overnight and then co-incubated with LY30. The indicator DCFDA is strongly activated by hydrogen peroxide and not by superoxide, but DCFDA can also be activated by highly reactive species such as hydroxyl radical. Quantification of the specificity has been published (Molecular Probes, The Handbook, Invitrogen).